2. Learning Analytics of Students' Academic Performance
Recently, learning analytics has been used to disclose patterns that exert an impact on student learning. Specifically, Van Leeuwen et al.
[14] have used learning analytics tools in a computer-supported collaborative learning environment to motivate and guide teachers in providing better interventions and in supporting collaborative groups of students faced with problems regarding cognitive activities. Moreover, the search for successful patterns for timely interventions led Sousa-Vieira et al.
[15] to conduct an in-depth examination of student activities on the SocialWire platform. Particularly, this platform programs three types of online activities: (a) pre-class activities, (b) questionnaires before partial exams, and (c) the use of forums for collaborative learning. Comparing the results obtained through various success/failure classifiers, the
resea
uthorchers concluded that the student final course grades are best predicted with the pace of the activities in which they participated, that is, the number of events per unit of time, instead of the type of initial activity. Moreover, Teo et al.
[16] demonstrated the usefulness of learning analytics methods in analyzing knowledge creation and collaboration in an online electric and electronics engineering course, whereas Kim et al.
[17] used learning analytics to support self-regulated learning in asynchronous online courses.
Undoubtedly, technological advances have improved the design and development processes of educational applications. In addition, interest in the use of ICT to enhance and predict academic performance has emerged
[18][19][20]. Some studies have focused on identifying hidden knowledge and patterns using data mining techniques
[21]. As such, applications and systems have experienced exponential growth in recent years in this field.
Pandey and Taruna
[22] developed multiple classifiers using K-nearest neighbor, and decision trees to predict academic performance. The
resea
uthorchers used a data-set on academic information as well as demographic information from a university in India to predict the academic performance of undergraduate engineering students. The
authoresearchers mentioned that the proposed method can also be used for the development of decision support systems.
Hasan et al.
[23] used decision tree algorithms to achieve the prediction of academic performance. To test their methodology, records from 22 students that contained academic information and activities in Moodle were used. A mining tool, named the Waikato Environment for Knowledge Analysis and developed at the University of Waikato, New Zealand, was used to evaluate the decision tree algorithm along with access time in Moodle. The
resea
uthorchers found that the random forest tree approach obtained better results in this task than comparative decision tree algorithms. Similarly, Hamsa et al.
[24] also used decision trees along with their implemented genetic fuzzy systems and Fuzzy Fitness Finder. The
resea
uthorchers reported that the results obtained from the decision tree classifier enabled the lecturers to take better care of students. Alternatively, the fuzzy logic approach provided friendlier results, which provided students with mental satisfaction, whereas lecturers could attend them indirectly.
In the same area, Bravo-Agapito et al.
[25] examined the use of exploratory factor analysis, multiple linear regressions, cluster analysis, and correlation to determine whether students are engaged in the course and to predict their academic performance. The
resea
uthorchers used data from Moodle interaction, characteristics, and grades of 802 undergraduate students and found that the prediction of academic performance is principally based on four factors, namely, access (variables related to student access to Moodle, including visits to forums and glossaries), questionnaires (visits to and attempts to complete questionnaires), tasks (variables related to consulted and submitted tasks), and age. Moreover, the
authoresearchers reported that the age factor predicts that academic performance is inversely related to age.
Trujillo-Torres et al.
[26] focused on mathematical competence. They proposed that the perception of students, the relationship between teacher and students, the classroom, gender, teaching-learning methods, and motivation are crucial factors for achieving optimal academic performance. The study intended to determine the optimal algorithm model for predicting the maximum learning gain of students. They employed a 14-item questionnaire, which was validated using the Kaiser–Guttman criterion and Tucker–Lewis Index. The cross-sectional study recruited a total of 2018 high-school students. The results indicated that the role of the classroom and the teacher–student relationship exerted a large influence on mathematics scores. Along a similar research line, Sharabiani et al.
[27] designed a prediction model using Bayesian networks to forecast the grades of engineering students in three courses. The study examined the records of 300 students to test the proposed model and used 10 variables, such as demographic data and scores obtained from previous courses. The accuracy exhibited by their approach was compared with other models, such as decision trees, K-nearest neighbors, and naive Bayes. In this direction, D’Uggento et al.
[28] also identified the usefulness of adopting a periodic monitoring system, which considers statistical techniques, such as logistic regression, survival analysis, and Cox regression model. These techniques enabled the early detection and modification of factors to achieve optimal results regarding students’ expectations and quality of higher education. The
resea
uthorchers used data from 7485 freshmen students enrolled in an academic year.
In the search for factors that exert various impacts on learning, Akhtar et al.
[29] used a computer support collaborative learning environment in a computer laboratory course to monitor student participation and to predict student success. The
resea
uthorchers found that achievement was positively correlated with course attendance, grouping with peers, and time allocation for task, whereas it was negatively correlated with the seating distance of students relative to the position of the lecturer. Using the linear regression approach, the
authoresearchers suggested that learning analytics can be used to predict academic performance and to identify students at risk of course failure. Similarly, Atkinson
[30] investigated the relationship of learning style, gender, and prior experience in design and technology among trainee teachers in their degree program. Although the results from the learning style groupings (verbal-visual and holistic-analytic) did not meet expectations and, although the conclusions about gender differences lacked a consensus, the study observed a positive relationship between achievement and past experience.
Regarding the role of anxiety on the learning outcomes of students, Chapell et al.
[31] investigated the relationship between test anxiety and academic performance on a large sample composed of 4000 undergraduate and 1414 graduate students enrolled in public universities in the USA enrolled in different majors. Using descriptive statistics, the
resea
uthorchers observed a small but significant inverse relationship between these two variables. Moreover, Vitasari et al.
[32] investigated the relationship between study anxiety and academic performance on a large sample of engineering students in Malaysia. The results demonstrated a significant correlation between high levels of anxiety and low levels of academic performance. Furthermore, the study concluded that anxiety during studying is a major predictor of academic performance and exerts a detrimental effect on student academic achievement. In similar research, Balogun et al.
[33] scrutinized the moderating role of achievement motivation in the relationship between test anxiety and academic performance among undergraduate students in Nigeria. The results indicated that, although test anxiety and achievement motivation exerted negative and positive effects on academic performance, respectively, achievement motivation significantly moderated these relationships. Therefore, the
resea
uthorchers concluded that universities should design appropriate psycho-educational interventions to enhance the achievement motivation of students.
Nowadays, evaluation should be aligned with specific competencies, such that students can exhibit their understanding and abilities through examinations so that teachers can improve their teaching
[34]. Empirical evidence illustrates that an active learning environment encourages students to be more open and committed. When evaluation considers class participation, quizzes, lab experiments, and presentations, in addition to written exams, then students obtain a better well-rounded view of their capabilities.